Model-based Automatic Segmentation of Ascending Aorta from Multimodality Medical Data

Abstract

Automatic Ascending Aorta Segmentation is one of the important steps towards automatic segmentation of the whole cardiac tree. This paper presents a novel approach for the automatic segmentation of the ascending aorta from two imaging modalities: CTA (Computed Tomography Angiography) and PC-MRI (Phase-Contrast Magnetic Resonance Images). The novel approach is an algorithm that works without the need for setting manual seed points or applying preprocessing steps or setting a region of interest. Instead, the proposed algorithm automatically detects and segments the ascending aorta using an ascending aorta model built from its anatomical features. The proposed segmentation algorithm begins with aorta detection through features model fitting augmented with Hough transform, where the ascending aorta is identified from the descending aorta and any other circular structures based on the proposed model. After detection, the whole ascending aorta is segmented up from the aortic arch down to the ostia points using a novel automatic seeded region growing algorithm. The proposed algorithm is fully automatic, works in real-time and robust as parameters used are the same for all the tested datasets. The detection and segmentation of the ascending aorta succeeded in all test cases acquired from the two imaging modalities; proving the robustness of the proposed ascending aorta model and algorithm for the automatic segmentation process even on data from different modalities and different scanner types. The accuracy of the segmentation has a mean Dice Similarity Coefficient (DSC) of 94.72% for CTA datasets and 97.13% for PC-MRI datasets.